diff --git a/app/benchmark_fft/main.cpp b/app/benchmark_fft/main.cpp
index aa21658..6b41c52 100644
--- a/app/benchmark_fft/main.cpp
+++ b/app/benchmark_fft/main.cpp
@@ -80,7 +80,7 @@ int main() {
   pls::internal::helpers::run_mini_benchmark([&] {
     complex_vector input = initial_input;
     fft(input.begin(), input.size());
-  }, 8, 1000);
+  }, 7, 30000);
 
   PROFILE_SAVE("test_profile.prof")
 }
diff --git a/lib/pls/include/pls/internal/base/system_details.h b/lib/pls/include/pls/internal/base/system_details.h
index 17b51d7..01543c0 100644
--- a/lib/pls/include/pls/internal/base/system_details.h
+++ b/lib/pls/include/pls/internal/base/system_details.h
@@ -2,10 +2,19 @@
 #ifndef PLS_SYSTEM_DETAILS_H
 #define PLS_SYSTEM_DETAILS_H
 
-#include <cstdint>
-#if (COMPILER == MVCC)
-#include <emmintrin.h>
+#if MSVC
+#ifdef _M_X86
+#define ARCH_X86
+#endif
+#endif
+
+#if GCC
+#ifdef __i386__
+#define ARCH_X86
 #endif
+#endif
+
+#include <cstdint>
 
 namespace pls {
 namespace internal {
@@ -53,7 +62,9 @@ constexpr pointer_t CACHE_LINE_SIZE = 64;
  * Choose the implementation appropriate for your compiler-cpu combination.
  */
 inline void relax_cpu() {
+#ifdef ARCH_X86
   asm volatile("pause":: : "memory");
+#endif
 }
 
 }